Polarimetric SAR change detection method, device, equipment and medium

Abstract

The present invention discloses a polarimetric SAR change detection method, device, equipment and medium, which relate to the field of radar image processing technology. The method acquires a time-series PolSAR image, uses JBLD divergence to calculate the similarity measure of the multi-phase covariance matrix; calculates a time-series edge intensity map based on the similarity measure; uses the time-series edge intensity map to initialize the cluster center and introduces a dynamic edge constraint mechanism to suppress superpixels from crossing the image edge during the iteration process, and outputs the superpixel segmentation result; constructs an image topology representation that integrates time-series feature similarity, spatial adjacency and cross-phase cross-feature similarity; constructs an energy function containing node cost and edge cost, solves the energy minimization problem through quadratic pseudo-Boolean optimization, and obtains a change detection map. The present invention can avoid errors caused by regional discontinuity and boundary fuzziness in superpixel segmentation, and exhibits strong robustness in both natural objects and complex urban building scenes.

Description

Technical Field

[0001] The present invention relates to the field of radar image processing technology, and in particular to a polarimetric SAR change detection method, device, equipment and medium. Background Art

[0002] Synthetic Aperture Radar (SAR) enables high-resolution, all-weather, all-day Earth observation by actively transmitting microwave signals and receiving backscattered waves from ground objects. It is widely used in fields such as topographic mapping and disaster monitoring. Traditional SAR acquires target scattering intensity information through a single polarization channel, but its ability to analyze complex ground objects is limited. Polarimetric SAR (PolSAR) further expands on this by acquiring target scattering matrix information through full polarization observations. This allows for a more comprehensive analysis of ground object characteristics such as geometric structure, dielectric constant, and scattering properties, significantly improving the accuracy of ground object classification and dynamic monitoring. Change detection, a key task in PolSAR image analysis, aims to identify surface changes occurring in the same area over different periods of time. However, its performance is severely limited by factors such as the high-dimensional polarization scattering characteristics of the data, coherent speckle noise, and multi-temporal radiation differences.

[0003] Existing PolSAR image change detection methods fall into two main categories: pixel-level methods and region-level methods. Pixel-level methods (such as those based on statistical information theory and hypothesis testing theory) directly compare differences in the polarization covariance matrix of individual pixels. While these methods can capture subtle changes, they are highly sensitive to noise, prone to false alarms and missed detections, and lack spatial consistency constraints, resulting in fragmented regions of change. Regional-level methods aggregate pixels into superpixel units and utilize contextual information to suppress noise interference, but they still face the following bottlenecks: 1. Insufficient temporal information fusion: Existing temporal superpixel generation algorithms fail to effectively capture the joint statistical laws of multi-temporal scattering characteristics, easily leading to cross-temporal boundary misalignment in areas of gradual or weak scattering changes, weakening segmentation consistency; 2. Lack of edge constraint mechanisms: Existing superpixel methods only use edge information to initialize cluster centers and do not dynamically suppress cross-edge misclassification during the iteration process, causing superpixels to cross real-world object boundaries and undermining the structural integrity of the object targets; 3. Inefficient utilization of topological structure information: Regional comparison-based methods rely on direct feature differences and lack modeling of image space and feature topological relationships. This leads to sensitivity to multi-temporal radiation differences and insufficient detection capabilities for areas of weak scattering changes. Furthermore, existing supervised deep learning-based methods are limited by the representativeness and coverage of training samples in complex scenes (such as urban building changes), resulting in significantly reduced generalization performance. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes a polarimetric SAR change detection method, device, equipment and medium. The edge-constrained temporal PolSAR image superpixel segmentation is to highlight the differences in the time phases with significant changes by defining a new temporal polarimetric similarity measure, and to introduce an edge constraint mechanism to suppress the problem of inaccurate segmentation caused by superpixels crossing the image edge, thereby significantly improving the fitting accuracy of superpixels to the actual ground object boundaries. The PolSAR image change detection based on graph structure energy optimization uses superpixels as the basic processing unit to construct an image topological representation that integrates temporal feature similarity and spatial adjacency. Then, the ability to discriminate weak scattering change areas is enhanced by measuring the cross-temporal feature similarity, and combined with the global energy model, noise suppression and maintenance of the integrity of the change area are achieved.

[0005] In order to achieve the above object, the technical solution of the present invention is as follows:

[0006] A polarimetric SAR change detection method comprises the following steps:

[0007] Acquire time-series PolSAR images and use JBLD divergence to calculate the similarity metric of the multi-temporal covariance matrix. Calculate the time-series edge intensity map based on the similarity metric to obtain the time-series edge intensity map. Use the time-series edge intensity map to initialize the cluster center and introduce a dynamic edge constraint mechanism during the iteration process to suppress superpixels from crossing the image edge. Output the superpixel segmentation result.

[0008] Based on the superpixel segmentation results, an image topological representation is constructed that integrates temporal feature similarity, spatial adjacency, and cross-temporal feature similarity. Based on the image topological representation, an energy function including node cost and edge cost is constructed. The energy minimization problem is solved through quadratic pseudo-Boolean optimization to obtain the change detection map of the dual-phase PolSAR image.

[0009] Preferably, the similarity measurement of multi-phase covariance matrices using JBLD divergence includes the following steps:

[0010] The functional expression of the JBLD divergence is: (1)

[0011] in, C 1, C 2 represents the two polarization covariance matrices, || represents the determinant of the matrix, and log(·) represents the natural logarithm;

[0012] Based on the statistical independence characteristics and maximum criterion between multi-temporal PolSAR images, the temporal distance metric is constructed as follows: (2)

[0013] in, C i,j Indicates thei Phase j The covariance matrix of the region, Z 1 and Z 2 represents the temporal joint covariance matrix of the two regions, N Indicates the number of phases.

[0014] Preferably, calculating the temporal edge strength based on the similarity metric to obtain a temporal edge strength graph includes the following steps:

[0015] A Gaussian filter is used to assign higher weights to pixels closer to the center. The horizontal Gaussian window function expression is: (3)

[0016] in,( x , y ) is the relative coordinate value, and through s x and s y Control the speed of window weight change in two directions respectively;

[0017] For a single-phase PolSAR image, the average covariance matrix of the areas on both sides of the center line is calculated as: (4)

[0018] in, i f is the direction angle of the center line, and ,Ω k , k =1,2 represents the set of pixels on both sides of the center line;

[0019] Combining equations (2) and (4), we can get the temporal edge intensity map, which is expressed as: (5)

[0020] in, and Indicates the i The bilateral average covariance matrix is ​​calculated using formula (4) under each time phase.

[0021] Preferably, a dynamic edge constraint mechanism is introduced in the iterative process to suppress superpixels from crossing the image edge, comprising the following steps:

[0022] Based on the time series edge strength graph, the edge constraint term is determined, and the expression is: (6)

[0023] Among them, EDGE norm is the normalized edge intensity map, line( i,j ) represents any two pixels i and j The line segments between p Represents pixels on a line segment;

[0024] SLIC clustering similarity measure for time-series PolSAR images D SLIC The overall structure is divided into three parts: feature similarity, edge constraint, and spatial similarity, which are defined as follows: (7)

[0025] in, α and β are the coefficients of edge constraint and spatial similarity weight respectively, S is the grid sampling step calculated based on the number of superpixels, D P is the maximum absolute percentage power similarity, D S is spatial similarity.

[0026] Preferably, a post-processing optimization process is also included, including the following steps:

[0027] Through connected region analysis, isolated superpixel blocks with an area smaller than a preset threshold are merged to eliminate over-segmentation noise;

[0028] Morphological closing operations are used to smooth superpixel edges and fill in small breaks caused by local similarity fluctuations.

[0029] Preferably, based on the superpixel segmentation results, constructing an image topology representation that integrates temporal feature similarity, spatial adjacency, and cross-temporal feature similarity includes the following steps:

[0030] The polarization scattering features of the PolSAR image are extracted using the polarization target decomposition method and globally normalized. The average polarization scattering features of the superpixel block are used as the feature vector of the node. The nodes of the constructed graph are: (10)

[0031] Among them, the node , and Respectively represent the first i The average polarization scattering characteristics of superpixels, N S represents the number of superpixels;

[0032] Constructing graph edges based on spatial similarity and feature similarity e , the expression is: (11)

[0033] in, and are edges constructed based on feature similarity in two phases, e S It is an edge built based on spatial similarity and is defined as follows:

[0034] (12)

[0035] (13)

[0036] (15)

[0037] in, express t In the image at phase 1 i The K nearest neighbor set of superpixel feature similarity; d s ( V i ,V j ) represents the spatial distance between the geometric centers of two superpixels.

[0038] Preferably, the expression of the energy function is: (16)

[0039] in, i const is a fixed offset value in the energy function, set to 0, i i ( L i ) represents a node V i In label L i The price of time, i ij ( L i ,L j ) indicates that the superpixel pairs with feature similarity or spatial proximity are in the label ( L i ,L j ) when the cost.

[0040] Based on the above content, the present invention also discloses a polarimetric SAR change detection device, comprising:

[0041] The superpixel segmentation module is used to acquire time-series PolSAR images and use the JBLD divergence to measure the similarity of multi-phase covariance matrices. Based on the similarity measure, the time-series edge intensity map is calculated to obtain the time-series edge intensity map. The time-series edge intensity map is used to initialize the cluster center and a dynamic edge constraint mechanism is introduced in the iterative process to suppress superpixels from crossing the image edge. The superpixel segmentation result is output.

[0042] The change detection module is used to construct an image topology representation based on the superpixel segmentation results, which integrates temporal feature similarity, spatial adjacency, and cross-temporal feature similarity. Based on the image topology representation, an energy function containing node costs and edge costs is constructed, and the energy minimization problem is solved through quadratic pseudo-Boolean optimization to obtain a change detection map of the dual-phase PolSAR image.

[0043] Based on the above content, the present invention further discloses a computer device, comprising: a memory for storing a computer program; and a processor for implementing any of the above methods when executing the computer program.

[0044] Based on the above content, the present invention further discloses a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, any of the above methods is implemented.

[0045] Based on the above technical solution, the beneficial effects of the present invention are:

[0046] 1) This paper proposes a Jensen-Bregman logarithmic determinant (Jensen-BregmanLog Det, JBLD) divergence similarity metric based on the maximum criterion. This metric adaptively focuses on the difference information of significantly changing phases, suppresses the noise interference of weakly fluctuating phases, and significantly improves the accuracy of time series segmentation.

[0047] 2) A dynamic edge constraint mechanism embeds a normalized edge strength map into the Simple Linear Iterative Clustering (SLIC) similarity metric. By penalizing pixel assignments that cross strong edges, this ensures that superpixels strictly adhere to the actual object boundaries, effectively maintaining the continuity and structural integrity of the object outline.

[0048] 3) This method uses superpixels as nodes to construct a graph structure representation that integrates temporal feature similarity and spatial adjacency. This structure representation innovatively introduces a cross-temporal feature similarity metric. By jointly analyzing the topological correlations between nodes across time, it significantly enhances the ability to discriminate weak scattering changes, demonstrating strong robustness in both natural features and complex urban building scenes.

[0049] 4) The present invention provides a change detection framework based on graph structure energy optimization. This detection framework constructs an energy function that includes node cost (a prior penalty term based on change sparsity) and edge cost (a state consistency constraint based on topological association). It uses the Quadratic Pseudo-Boolean Optimization (QPBO) algorithm to solve the energy minimization problem and achieves efficient global convergence through the maximum flow / minimum cut theory, avoiding the regional discontinuity and boundary fuzziness problems caused by traditional threshold segmentation methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a flow chart of a polarimetric SAR change detection method according to an embodiment;

[0051] Figure 2 The figure is a schematic diagram of feature similarity between superpixels in a polarization SAR change detection method in an embodiment. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0053] like Figure 1 As shown, this embodiment provides a polarimetric SAR change detection method, which includes two parts: edge-constrained time-series PolSAR image superpixel segmentation and graph structure energy optimization-based PolSAR image change detection. Specific details are as follows:

[0054] 1. Superpixel segmentation of time-series PolSAR images based on edge constraints

[0055] 1.1. Using JBLD divergence to calculate the similarity measure of multi-temporal covariance matrices (similarity measure is used in the calculation of edge maps and SLIC clustering process)

[0056] While significantly improving computational efficiency, the JBLD divergence maintains comparable performance in terms of accuracy and robustness compared to the symmetric revised Wishart distance (SRW) metric. It is defined as: (1)

[0057] in, C 1, C 2 represents the two polarization covariance matrices, || represents the determinant of the matrix, and log(·) represents the natural logarithm. Based on the statistical independence characteristics and maximum criterion between multi-temporal PolSAR images, the temporal distance metric constructed in this embodiment is as follows: (2)

[0058] in, C i,j Indicates the i Phase j The covariance matrix of the region, Z 1 and Z 2 represents the temporal joint covariance matrix of the two regions, N Indicates the number of phases.

[0059] 1.2. Calculation of temporal edge intensity map based on Gaussian filter

[0060] In the edge detection framework based on the bilateral window filter, the Gaussian window can avoid the false local maximum caused by strong noise compared to the rectangular window, and thus can generate more robust edges. The Gaussian filter assigns higher weights to pixels close to the center. The horizontal Gaussian window function is defined as: (3)

[0061] in,( x , y ) is the relative coordinate value, and through s x and s y Control the weight change speed of the window in two directions respectively. For a single-phase PolSAR image, the average covariance matrix of the area on both sides of the center line is calculated by the following formula: (4)

[0062] in, i f is the direction angle of the center line, and Ω k , k =1,2 represents the set of pixels on both sides of the center line. Combining formula (2) and formula (4), the edge strength calculation formula of the time series PolSAR image is finally obtained: (5)

[0063] in, and Indicates the i The bilateral average covariance matrix is ​​calculated using formula (4) in each time phase. The temporal edge strength map EDGE can be obtained by traversing the entire image using formula (5).

[0064] 1.3. Initialize superpixel cluster centers

[0065] The initial cluster centers are sampled in a regular grid SEvenly spread over the entire image, and then move the cluster center to the lowest edge intensity in the 3×3 neighborhood in the temporal edge intensity graph EDGE.

[0066] 1.4 K-means clustering based on dynamic edge constraint mechanism

[0067] This embodiment proposes a dynamic edge constraint mechanism that integrates edge information into the SLIC clustering similarity metric and establishes an edge-oriented similarity metric criterion, so that the generated superpixels can accurately fit the image edge features, thereby significantly improving the segmentation accuracy of potential targets. The edge constraint term is defined as follows: (6)

[0068] Among them, EDGE norm is the normalized edge intensity map, line ( i,j ) represents any two pixels i and j The line segments between p Represents the pixels on the line segment. If two pixels do not cross the edge of the image, the edge constraint term will be low. Conversely, the edge constraint term between two pixels that cross the edge of the image will be high.

[0069] Using this property, the SLIC clustering similarity metric of time-series PolSAR images is D SLIC The overall structure is divided into three parts: feature similarity, edge constraint, and spatial similarity, which are defined as follows: (7)

[0070] in, α and β are the coefficients of edge constraint and spatial similarity weight respectively, S is the grid sampling step calculated based on the number of superpixels. Feature similarity is dominated by the similarity measure defined in formula (2), which can maximize the use of the difference information of the phase with the largest fluctuation while improving the computational efficiency. D P is the maximum absolute percentage power similarity, defined as: (8)

[0071] in, P i,1 and P i,2 Respectively represent i The total scattered power of two points in the time phase is D P The value range is [0,1], (1+ D P) provides a maximum gain of two times for feature similarity. Finally, spatial similarity is used to improve the compactness of generated superpixels: (9)

[0072] in,( x 1, y 1) and ( x 2, y 2) Represents the spatial coordinates of two points.

[0073] 1.5 Post-processing optimization

[0074] After edge-constrained SLIC clustering, post-processing optimization is performed on the initial superpixel segmentation results. First, connected component analysis is used to merge isolated superpixel blocks with an area smaller than a preset threshold to eliminate over-segmentation noise. Second, morphological closing operations are used to smooth superpixel edges and fill in minor breaks caused by fluctuations in local similarity.

[0075] 2. PolSAR Image Change Detection Based on Graph Structure Energy Optimization

[0076] 2.1 Superpixel Feature Extraction

[0077] To construct a graph representing the topological structure of PolSAR images G ={ v , e First, the polarization target decomposition method is used to extract the polarization scattering characteristics of the PolSAR image. F (For example, using the Yamaguchi decomposition method to obtain F ), and perform global normalization, then use the average polarization scattering characteristics of the superpixel block as the feature vector of the node, and then construct the node of the graph as: (10)

[0078] in, , and Respectively represent the first i The average polarization scattering characteristics of superpixels, N S Indicates the number of superpixels, and this feature will be used to refer to its corresponding superpixel.

[0079] 2.2. Topological adjacency relationship construction

[0080] The topological adjacency relationship of an image is also called the graph G edge e Constructed based on spatial similarity and feature similarity: (11)

[0081] in, and They are the edges constructed based on feature similarity in two phases. e , defined as follows:

[0082] (12)

[0083] (13)

[0084] in, express t In the image at phase 1 i The K nearest neighbor set of superpixel feature similarity defines any two superpixels and The feature similarity between them is: (14)

[0085] in, Indicates the division of corresponding positions of vectors. t 1st moment i superpixels, using formula (14) and t At time 1, find the feature similarity of other superpixels and sort them, and take the top The minimum value can be obtained I will not go into details here. t 2. The situation at moment 2. e S It is an edge built based on spatial similarity and is defined as follows: (15)

[0086] in, d s ( V i ,V j ) represents the spatial distance between the geometric centers of two superpixels, S is the superpixel grid sampling step.

[0087] 2.3 Energy function construction

[0088] Based on the topological structure representation of the dual-phase PolSAR image, the PolSAR image change detection problem is converted into an energy minimization problem, and its energy function is: (16)

[0089] in, i const is a fixed offset value in the energy function, set to 0, i i ( L i ) represents a node Vi In label L i The price of time, i ij ( L i ,L j ) indicates that the superpixel pairs (i.e., the edges of the graph) with feature similarity or spatial proximity are in the label ( L i ,L j ). Therefore, the key to the change detection method based on energy minimization is the design of node cost and edge cost functions.

[0090] For this reason, Figure 2 As shown, this embodiment first defines any two nodes as and There are 3 categories and 6 feature similarities between the four superpixels, namely, intra-node feature similarity and , feature similarity between nodes and , and cross-feature similarity and .

[0091] Based on these six feature similarities, four edge cost functions are constructed as follows:

[0092] (17)

[0093] (18)

[0094] (19)

[0095] (20)

[0096] Based on the fact that only a small part of the area changes while most of the area remains unchanged in the actual change detection problem, a sparse penalty term based on the change prior is designed, that is, the node cost function is: (twenty one)

[0097] In order to balance the impact of the node cost function and the edge cost function on the minimization process, this embodiment normalizes the four edge cost functions to obtain the normalized edge cost function : (twenty two)

[0098] in, The state of the edge is ( Li ,L j ), the gain coefficient can be normalized to ensure that the cost of the edge and the cost of the node are in the same order of magnitude.

[0099] 2.4 Energy Optimization Solution

[0100] The QPBO method is used to solve the minimization problem in Equation (16). The QPBO method is a maximum flow / minimum cut algorithm with excellent time complexity in graph optimization. It can efficiently solve large-scale problems and ensures that the results are globally optimal. Through this optimization process, the change state (changed / unchanged) of each superpixel is accurately determined, and a complete change detection map of the dual-phase PolSAR image is generated based on this. This effectively avoids the regional discontinuity and boundary fuzziness caused by traditional threshold segmentation.

[0101] It should be understood that, although the various steps in the above flow chart are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the above flow chart may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0102] In one embodiment, a polarimetric SAR change detection device is provided, comprising:

[0103] The segmentation module is used to acquire time-series PolSAR images and calculate the similarity metric of the multi-phase covariance matrix using the JBLD divergence. The time-series edge intensity map is calculated based on the similarity metric to obtain the time-series edge intensity map. The time-series edge intensity map is used to initialize the cluster center and a dynamic edge constraint mechanism is introduced in the iterative process to suppress superpixels from crossing the image edge. The superpixel segmentation result is output.

[0104] The change detection module is used to construct an image topology representation based on the superpixel segmentation results, which integrates temporal feature similarity, spatial adjacency, and cross-temporal feature similarity. Based on the image topology representation, an energy function containing node costs and edge costs is constructed, and the energy minimization problem is solved through quadratic pseudo-Boolean optimization to obtain a change detection map of the dual-phase PolSAR image.

[0105] The devices and modules described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0106] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor performs the following steps when executing a computer program:

[0107] Acquire time-series PolSAR images and use JBLD divergence to calculate the similarity metric of the multi-temporal covariance matrix. Calculate the time-series edge intensity map based on the similarity metric to obtain the time-series edge intensity map. Use the time-series edge intensity map to initialize the cluster center and introduce a dynamic edge constraint mechanism during the iteration process to suppress superpixels from crossing the image edge. Output the superpixel segmentation result.

[0108] Based on the superpixel segmentation results, an image topological representation is constructed that integrates temporal feature similarity, spatial adjacency, and cross-temporal feature similarity. Based on the image topological representation, an energy function including node cost and edge cost is constructed. The energy minimization problem is solved through quadratic pseudo-Boolean optimization to obtain the change detection map of the dual-phase PolSAR image.

[0109] In one embodiment, a storage medium storing computer-readable instructions is further provided. When the computer-readable instructions are executed by one or more processors, the one or more processors implement the following steps:

[0110] Acquire time-series PolSAR images and use JBLD divergence to calculate the similarity metric of the multi-temporal covariance matrix. Calculate the time-series edge intensity map based on the similarity metric to obtain the time-series edge intensity map. Use the time-series edge intensity map to initialize the cluster center and introduce a dynamic edge constraint mechanism during the iteration process to suppress superpixels from crossing the image edge. Output the superpixel segmentation result.

[0111] Based on the superpixel segmentation results, an image topological representation is constructed that integrates temporal feature similarity, spatial adjacency, and cross-temporal feature similarity. Based on the image topological representation, an energy function including node cost and edge cost is constructed. The energy minimization problem is solved through quadratic pseudo-Boolean optimization to obtain the change detection map of the dual-phase PolSAR image.

[0112] Computer-readable instruction storage media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0113] Each embodiment in this specification is described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences from other embodiments. In particular, the system embodiment is generally similar to the method embodiment, so the description is relatively simple. For relevant parts, refer to the description of the method embodiment.

Claims

1. A polarimetric SAR change detection method, characterized in that: The steps include: Acquire time-series PolSAR images and use JBLD divergence to calculate the similarity metric of the multi-temporal covariance matrix. Calculate the time-series edge intensity map based on the similarity metric to obtain the time-series edge intensity map. Use the time-series edge intensity map to initialize the cluster center and introduce a dynamic edge constraint mechanism during the iteration process to suppress superpixels from crossing the image edge. Output the superpixel segmentation result. Based on the superpixel segmentation results, an image topology representation is constructed that integrates temporal feature similarity, spatial adjacency, and cross-temporal feature similarity. Based on the image topology representation, an energy function including node cost and edge cost is constructed. The energy minimization problem is solved by quadratic pseudo-Boolean optimization to obtain the change detection map of the dual-phase PolSAR image, where: The method of constructing an image topology representation based on superpixel segmentation results that integrates temporal feature similarity, spatial adjacency, and cross-temporal feature similarity includes the following steps: The polarization scattering features of the PolSAR image are extracted using the polarization target decomposition method and globally normalized. The average polarization scattering features of the superpixel block are used as the feature vector of the node. The nodes of the constructed graph are: (10) Among them, the node , and represents the average polarization scattering characteristics of the i-th superpixel in the two phases, N S represents the number of superpixels; The edge ε of the graph is constructed based on spatial similarity and feature similarity, and the expression is: (11) in, and are edges constructed based on feature similarity in two phases, ε S It is an edge built based on spatial similarity and is defined as follows: (12) (13) (15) in, represents the K nearest neighbor set of the feature similarity of the i-th superpixel in the image at time t1; d s (V i ,V j ) represents the spatial distance between the geometric centers of two superpixels.

2. The polarimetric SAR change detection method according to claim 1, wherein: The similarity measurement of multi-temporal covariance matrices using JBLD divergence includes the following steps: The functional expression of the JBLD divergence is: (1) Where C1 and C2 represent two polarization covariance matrices, || represents the determinant of the matrix, and log(·) represents the natural logarithm; Based on the statistical independence characteristics and maximum criterion between multi-temporal PolSAR images, the temporal distance metric is constructed as follows: (2) Among them, C i,j represents the covariance matrix of the jth region in the i-th phase, Z1 and Z2 represent the temporal joint covariance matrices of the two regions, and N represents the number of phases.

3. The polarimetric SAR change detection method according to claim 2, wherein: The temporal edge strength is calculated based on the similarity metric to obtain a temporal edge strength graph, which includes the following steps: A Gaussian filter is used to assign higher weights to pixels closer to the center. The horizontal Gaussian window function expression is: (3) Among them, (x, y) is the relative coordinate value, and through σ x and σ y Control the speed of window weight change in two directions respectively; For a single-phase PolSAR image, the average covariance matrix of the areas on both sides of the center line is calculated as: (4) Among them, θ f is the direction angle of the center line, and ,Ω k , k=1,2 represents the set of pixels on both sides of the center line; Combining equations (2) and (4), we can get the temporal edge intensity map, which is expressed as: (5) in, and It represents the bilateral average covariance matrix calculated by formula (4) in the i-th phase.

4. The polarimetric SAR change detection method according to claim 3, wherein: The dynamic edge constraint mechanism is introduced in the iterative process to suppress superpixels from crossing the image edge, including the following steps: Based on the time series edge strength graph, the edge constraint term is determined, and the expression is: (6) Among them, EDGE norm is a normalized edge intensity map, line(i,j) represents the line segment between any two pixels i and j, and p represents the pixel on the line segment; SLIC clustering similarity measure D for time series PolSAR images SLIC The overall structure is divided into three parts: feature similarity, edge constraint, and spatial similarity, which are defined as follows: (7) Among them, α and β are the coefficients of edge constraint term and spatial similarity weight respectively, S is the grid sampling step calculated according to the number of superpixels, and D P is the maximum absolute percentage power similarity, D S is spatial similarity.

5. The polarimetric SAR change detection method according to claim 1, wherein: It also includes the post-processing optimization process, including the following steps: Through connected region analysis, isolated superpixel blocks with an area smaller than a preset threshold are merged to eliminate over-segmentation noise; Morphological closing operations are used to smooth superpixel edges and fill in small breaks caused by local similarity fluctuations.

6. The polarimetric SAR change detection method according to claim 1, wherein: The expression of the energy function is: (16) Among them, θ const is a fixed offset value in the energy function, set to 0, θ i (L i ) represents node V i In label L i The cost of time, θ ij (L i ,L j ) indicates that the superpixel pairs with feature similarity or spatial proximity are in the label (L i ,L j ) when the cost.

7. A polarimetric SAR change detection device, characterized in that: include: The superpixel segmentation module is used to obtain time-series PolSAR images and use JBLD divergence to measure the similarity of multi-temporal covariance matrices; The temporal edge intensity map is calculated based on the similarity metric to obtain the temporal edge intensity map. The cluster center is initialized using the temporal edge intensity map and a dynamic edge constraint mechanism is introduced in the iterative process to suppress superpixels from crossing the image edge. The superpixel segmentation result is output. The change detection module is used to construct an image topological representation based on the superpixel segmentation results, integrating temporal feature similarity, spatial adjacency, and cross-temporal feature similarity; Based on the image topology representation, an energy function including node cost and edge cost is constructed. The energy minimization problem is solved by quadratic pseudo-Boolean optimization to obtain the change detection map of the dual-phase PolSAR image, where: The method of constructing an image topology representation based on superpixel segmentation results that integrates temporal feature similarity, spatial adjacency, and cross-temporal feature similarity includes the following steps: The polarization scattering features of the PolSAR image are extracted using the polarization target decomposition method and globally normalized. The average polarization scattering features of the superpixel block are used as the feature vector of the node. The nodes of the constructed graph are: (10) Among them, the node , and represents the average polarization scattering characteristics of the i-th superpixel in the two phases, N S represents the number of superpixels; The edge ε of the graph is constructed based on spatial similarity and feature similarity, and the expression is: (11) in, and are edges constructed based on feature similarity in two phases, ε S It is an edge built based on spatial similarity and is defined as follows: (12) (13) (15) in, represents the K nearest neighbor set of the feature similarity of the i-th superpixel in the image at time t1; d s (V i ,V j ) represents the spatial distance between the geometric centers of two superpixels.

8. A computer device, characterized in that: The device comprises: a memory for storing a computer program; A processor, configured to implement the method according to any one of claims 1 to 6 when executing the computer program.

9. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images